Per Bothner scripsit:

> This assumes that libraries for reading zip file entries can read
> "relatively".

You can't unzip with sequential access only: random access is required,
because there may be false positives for the central directory signature.

> (The zip specification says the central directory contains for each
> entry a "relative offset of local header", but seems to say it is
> relative from the *start* of the "first disk", which doesn't help.)

What you aren't told is that most unzip programs do in fact scan from
the beginning to look for the first local header, and internally adjust
all offsets accordingly.  Some older proprietary zip programs may not do
this.  See the unzipsfx man page.

Multivolume zip files are obsolete.  They were devised to make it easy
to split a zip file over many floppies or other small devices for backup
purposes.  WP says they are still used for sending large zip files by
email, but I doubt if anyone actually does this now.

--
John Cowan          http://www.ccil.org/~cowan        xxxxxx@ccil.org
        "You need a change: try Canada"  "You need a change: try China"
                --fortune cookies opened by a couple that I know